Google launches Gemini 4 Carbon model with unchanged introductory pricing
Google introduced Gemini 4 Carbon, a higher-performing AI model sold at the same introductory rates as its predecessor, and highlighted its cybersecurity strengths and benchmark results.
Google announced the Gemini 4 Carbon model two weeks after releasing Gemini 4 Argon, keeping the introductory price of $2 per million input tokens and $10 per million output tokens for an unspecified period. After that period, the company plans to raise the rates to $4 per million input tokens and $20 per million output tokens, matching the pricing of Anthropic's Claude Opus 5.5.
Koray Kavukcuoglu, senior vice president at Google DeepMind and chief AI architect, described Carbon as the company's strongest model to date, especially for cybersecurity and long-running tasks. He noted that internal tests showed more than a two-fold improvement in generating secure code compared with Argon, and that the Chrome team reported the model performed over ten times better than a competing large-frontier model in creating high-quality patches.
"Gemini 4 Argon and Carbon are now leading at the frontier," Sundar Pichai, CEO of Google and Alphabet, said in a statement accompanying the firm’s Q3 2026 earnings.
Benchmark results place Carbon at the top of the Artificial Analysis Intelligence Index v5 and second on Arena.ai’s Best Overall Agent leaderboard, behind Claude Opus 5.5. It also leads in SEC-bench Pro, ExploitGym, and CWE-bench, while ranking fourth on WebDev Arena.
Google plans to make Carbon available to a limited group of cybersecurity defenders this week, with broader access to follow through its Fairwind program. The company’s Q3 earnings report highlighted that its first-party model APIs processed nearly 32 billion tokens per minute, four times the volume recorded a year earlier.